An Overview of Multiple Classifier Systems Based on Generalized Additive Models

K.W. De Bock, K. Coussement, Davy Cielen

Research output: Chapter in Book/Conference paperChapter

Abstract

Multiple classifier systems combine the decisions from individual classifiers to obtain a more accurate classifier. Multiple classifier systems are also known as ensemble methods, committee of classifiers, and mixture of experts. Three popular ways of creating the individual classifiers for multiple classifiers systems are bagging, random subspace modeling, and boosting. This chapter reviews the GAMens family of multiple classifier systems that use general additive models as a base classifier. It shows that generalized additive model (GAM)‐based multiple classifiers serve as a good extension to the ensemble‐based literature because they are stronger predictors than a single GAM‐based classifier. The chapter introduces the theory behind GAMens classifiers, and then looks at real‐world applications and how GAM‐based multiple classifier systems compare to other popular algorithms. The GAMensPlus algorithm combines the training and prediction phases of GAMens with an explanation phase in which two heuristics are introduced to allow model interpretation.
Original languageEnglish
Title of host publicationEnsemble Classification Methods with Applications in R
EditorsEsteban Alfaro, Matías Gámez, Noelia García
Place of PublicationUSA
PublisherJohn Wiley & Sons
Pages175-186
Number of pages12
ISBN (Electronic)9781119421559
ISBN (Print)9781119421092
DOIs
Publication statusPublished - 2019
Externally publishedYes

Publication series

NameEnsemble Classification Methods with Applicationsin R

Fingerprint Dive into the research topics of 'An Overview of Multiple Classifier Systems Based on Generalized Additive Models'. Together they form a unique fingerprint.

Cite this